Spaces:
Runtime error
Runtime error
File size: 5,022 Bytes
87d6a85 ce54bdc 87d6a85 ce54bdc 87d6a85 ce54bdc 87d6a85 ce54bdc 87d6a85 ce54bdc 87d6a85 ce54bdc 87d6a85 ce54bdc 87d6a85 ce54bdc 87d6a85 ce54bdc 87d6a85 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 |
# Loading key libraries
import streamlit as st
import os
import pickle
import numpy as np
import pandas as pd
import re
from pathlib import Path
from PIL import Image
import matplotlib.pyplot as plt
import seaborn as sns
import requests
# get absolute path and goo two levels up
DIRPATH = DIRPATH = os.path.dirname(os.path.dirname(os.path.dirname(os.path.realpath(__file__))))
# get path for app data
app_data_path =os.path.join(DIRPATH,'dev', 'datasets', 'app_data', 'Grocery.csv.crdownload')
# set api endpoint
URL = 'https://bright1-sales-forecasting-api.hf.space'
API_ENDPOINT = '/predict'
# Setting the page configurations
st.set_page_config(page_title = "Prediction Forecasting", layout= "wide", initial_sidebar_state= "auto")
# Setting the page title
st.title("Grocery Store Forecasting Prediction")
# Load the saved data
df = pd.read_csv(app_data_path)
# src\app\images1.jpg
image1 = Image.open('src/app/images1.jpg')
image2 = Image.open('src/app/image 2.jpg')
def make_prediction(store_id, category_id, onpromotion, year,month, dayofmonth,
dayofweek, dayofyear,weekofyear, quarter, is_month_start, is_month_end,
is_quarter_start, is_quarter_end, is_year_start, is_year_end,
year_weekofyear,city, store_type, cluster):
parameters = {
'store_id':int(store_id),
'category_id':int(category_id),
'onpromotion' :int(onpromotion),
'year' : int(year),
'month' : int(month),
'dayofmonth' :int(dayofmonth),
'dayofweek' : int(dayofweek),
'dayofyear' : int(dayofyear),
'weekofyear' : int(weekofyear),
'quarter' : int(quarter),
'is_month_start' : int(is_month_start),
'is_month_end' : int(is_month_end),
'is_quarter_start' : int(is_quarter_start),
'is_quarter_end' : int(is_quarter_end),
'is_year_start' : int(is_year_start),
'is_year_end' : (is_year_end),
'year_weekofyear' : int(year_weekofyear),
'city' : city,
'store_type' : int(store_type),
'cluster': int(cluster),
}
response = requests.post(url=f'{URL}{API_ENDPOINT}', params=parameters)
sales_value = response.json()['sales']
sales_value = round(sales_value, 4)
return sales_value
st.image(image1, width = 700)
st.sidebar.markdown('User Input Details and Information')
store_id= st.sidebar.selectbox('store_id', options = sorted(list(df['store_id'].unique())))
category_id= st.sidebar.selectbox('categegory_id',options = sorted(list(df['category_id'].unique())))
onpromotion= st.sidebar.number_input('onpromotion', min_value= df["onpromotion"].min(), value= df["onpromotion"].min())
year = st.sidebar.selectbox('year', options = sorted(list(df['year'].unique())))
month = st.sidebar.selectbox('month', options = sorted(list(df['month'].unique())))
dayofmonth= st.sidebar.number_input('dayofmonth', min_value= df["dayofmonth"].min(), value= df["dayofmonth"].min())
dayofweek = st.sidebar.number_input('dayofweek', min_value= df["dayofweek"].min(), value= df["dayofweek"].min())
dayofyear = st.sidebar.number_input('dayofyear', min_value= df["dayofyear"].min(), value= df["dayofyear"].min())
weekofyear = st.sidebar.number_input('weekofyear', min_value= df["weekofyear"].min(), value= df["weekofyear"].min())
quarter = st.sidebar.number_input('quarter', min_value= df["quarter"].min(), value= df["quarter"].min())
is_month_start = st.sidebar.number_input('is_month_start', min_value= df["is_month_start"].min(), value= df["is_month_start"].min())
is_month_end = st.sidebar.number_input('is_month_end', min_value= df["is_month_end"].min(), value= df["is_month_end"].min())
is_quarter_start = st.sidebar.number_input('is_quarter_start', min_value= df["is_quarter_start"].min(), value= df["is_quarter_start"].min())
is_quarter_end = st.sidebar.number_input('is_quarter_end', min_value= df["is_quarter_end"].min(), value= df["is_quarter_end"].min())
is_year_start = st.sidebar.number_input('is_year_start', min_value= df["is_year_start"].min(), value= df["is_year_start"].min())
is_year_end = st.sidebar.number_input('is_year_end', min_value= df["is_year_end"].min(), value= df["is_year_end"].min())
year_weekofyear = st.sidebar.number_input('year_weekofyear', min_value= df["year_weekofyear"].min(), value= df["year_weekofyear"].min())
city = st.sidebar.selectbox("city:", options= sorted(set(df["city"])))
store_type= st.sidebar.number_input('type', min_value= df["type"].min(), value= df["type"].min())
cluster = st.sidebar.selectbox('cluster', options = sorted(list(df['cluster'].unique())))
# make prediction
sales_value = make_prediction(store_id, category_id, onpromotion, year,month, dayofmonth,
dayofweek, dayofyear,weekofyear, quarter, is_month_start, is_month_end,
is_quarter_start, is_quarter_end, is_year_start, is_year_end,
year_weekofyear,city, store_type, cluster)
# get predicted value
if st.button('Predict'):
st.success('The predicted target is ' + str(sales_value))
|